Debugging race conditions?

[Mulyadi Santosa]

Hi...

On Tue, May 26, 2009 at 3:57 PM, pradeep singh
<pradeep.rautela at gmail.com> wrote:
> Hi All,
>
> I am having a tough time debugging concurrency issues it seems.
> I find myself flabbergasted at debugging panic because of race issues.
>
> I would like to ask people here, if they have any advice on how to
> properly approach these specific set of panics?
> How do they do it?
> Are there any rules of thumbs which can help me?

OK, first thing that I could recall quickly is: do you really
understand the flow of the code? do you know the possibilities of what
might happen if several code path happen at the same time?

In debugging race condition, I think it's important either by trying
to reduce the threads number (if it's solely a multi threading
program), or...in a complex situation like in Linux kernel.... try to
intersect with as least triggers as possible i.e is it possible to
trigger the panic with just interrupt handler code triggered by 2 CPUs
instead of 4 or 8 CPUs

regards,

Mulyadi.